Classification of voice disorders in children with cochlear implantation and hearing aid using multiple classifier fusion

摘要

Speech production and speech phonetic features gradually improve in children by obtaining audio feedback after cochlear implantation or using hearing aid. In this study, voice disorders in children with cochlear implantation and hearing aid are classified. 30 Persian children participated in the study, including 6 children in levels 1 to 3 and 12 in level 4. Voice samples of 5 isolated Persian words “mashin”, “mar”, “moosh”, “gav” and “mouz” are analyzed. 4 level for the voice quality are considered, the higher the level the less the speech disorders. “Frame-based” and “word-based” features are extracted from speech signal. Some of the frame-based features include fundamental frequency, formants and nasality and word-based features include phase space features and wavelet coefficients. For frame-based features, hidden Markov models are used as classifiers and for word-based features, neural network is used. After Classifiers fusion with three methods: Majority Voting Rule, Linear Combiner and Stacked fusion, the best classification rates are obtained using frame-base and word-base features excluding third to second formant ratio and MVR rule (level1:100%, level2: 93.75%, level3: 100%, level4: 94%). Output of the study can help speech pathologists to follow up voice disorder recovery in children with cochlear implantation or hearing aid.